chezburger ballin

talking test coming in a bit

i got 2 phones

Another world record for THE MOST OVERTRAINED RVC MODEL TO *EVER* BE POSTED ON AI HUB

1million next

speaking

Then after that, 1 billion.

prolly gon do 150,000 then i'll stop, too fuckin much

can prolly make a 1 million one but i highly doubt the program will resist

Can you give us the loss graph for this model?

just as an example the loss graph bugged the living fuck out. it didnt exist.

so i think this is the limit

the program shat itself

can't cause the tensorboard litterally disappeared

don't know why but im guessing just *too much*

ya but idk which audio

"LOL" sound effect

oh yeah i saw that before you replied

1 million epoch model when

ion think the programs able to but... i'll test

incase i get a BSOD imma just assume this the LIMIT

fym this is undertrained asf

?

yea LMAOO

try that

amen it works

you being sarcastic bro?

give me credit for dataset tho

ok in like 2 hours

no

i am 100% serious

this model is undertrained

15000 would have probably worked better

bro how? how? how would a model that was trained for 100 THOUSAND EPOCHS be undertrained?

hmm

undertrained

its not enough

needs more

10000000

sounds pretty undertrained rn

peper drop the act bro it was funny the first time you said it was undertrained now it's just annoying💀

rude..

100k is crazyyy

how long did it take to train?

*days*.

jesus

i used applio and shrunk the dataset to an unfathomable small amount of nearly a kilobyte to get this

to make it as fast as possible

did u have to like

my electricity bill is gonna have a 800+ euro credit just for this

continue retraining cuz runtime closes after like

4 hours

at times yea

tho applio saves everything

so it wasnt too much of an issue

100,000 epochs

Do One With 999999 Epochs

i know how he did this

inspect element

can't wait for someone to do 100,001 epochs

chez chez chez 🔥🔥